How to Analyze Instagram Comments Without AI

Why start with counting

“Analyze the comments” usually means one of a few concrete questions: How many people engaged? What did they talk about? What did they ask? Who was involved? When did it happen? Every one of those can be answered by counting things that are actually in the text — words, tags, question marks, timestamps — without any model guessing at meaning.

Counting has real advantages. It is deterministic: run it twice and you get the same answer. It is transparent: if you report that “shipping” appeared in 84 comments, anyone can open the file and check. It handles every language equally well, which matters on accounts where commenters write in Spanish, Portuguese or Arabic. And it costs nothing.

The IGComments stats panel does this for any public post or Reel, for free and with no AI: paste a link on the home page, press Get comments, and open the stats. The comment analyzer page lists everything it shows. The sections below explain how to read each part — and how to reproduce it in a spreadsheet.

1. Size and shape of the conversation

Start with the headline numbers:

Total comments and unique commenters — how much was said, and by how many people. A big gap means repeat commenters or long conversations.
Repeat commenters (commented two or more times) and the most active commenters list — who drove the conversation.
Replies vs top-level comments — a high share of replies means people were talking to each other (or you were answering them); a low share means mostly one-off reactions.
Average comment length — short averages usually mean emoji and one-word reactions; long ones mean opinions, stories or questions.

These numbers become useful when you compare posts. One post’s “420 comments” means little alone; the fact that it had twice the unique commenters of your average post, with three times the replies, tells you something. See counting unique commenters for the details.

2. What people talked about: words, hashtags, mentions, emoji

Most used words. IGComments counts the words in each comment, ignoring common filler words in English, Spanish and Portuguese (“the”, “que”, “não”), so the list surfaces topics: product names, “price”, “colour”, “recipe”, “song”. Each word is counted once per comment, so one person repeating a word ten times doesn’t dominate.
Top hashtags. Usually campaign tags or community tags. Unexpected ones can reveal where viewers came from.
Most-mentioned accounts. On giveaways it is friends being tagged; on brand posts it is often people tagging a friend who would like the product, or tagging another brand they compare you to.
Top emoji. A crude but honest mood check: a list led by hearts and fire reads very differently from one led by laughing faces or eye-rolls.

Word counts need a human to interpret them. “Price” near the top could be enthusiasm (“great price!”) or complaints (“price is too high”). That is your cue to search for the word and read the comments behind it — see how to search Instagram comments.

3. What people asked

The Questions count is comments containing ? or ¿, and the Questions filter lists them. This is often the most valuable slice of any post: questions are explicit, actionable and easy to answer. Read them all if there are a few dozen; on large posts, combine the filter with a search (“size”, “ship”, “price”) to group them.

The rule is deliberately simple, so it misses questions written without a question mark (“how much” on its own) and includes rhetorical ones. That trade-off is the point: you know exactly what was counted. For a fuller method including buying-intent keywords, see finding questions and buying intent.

4. When it happened, and what landed

Comments per day shows the shape of the conversation over time (by UTC date). Most posts spike on day one and fade. A second spike usually means something happened: the post was shared, a Reel resurfaced, a story drove people back, or a giveaway deadline approached. Use the date range filter to read what people said during a spike.

Most-liked comments are the comments other viewers endorsed. They are frequently a better summary of audience opinion than any sentiment score: if the top comment is a joke, the post was received as entertainment; if it is a sharp question, other people wanted that answer too. Sorting by most liked in the search view shows the full ranking.

Doing the same analysis in a spreadsheet

Export the comments (CSV or Excel) and you can reproduce or extend any of this. Comment text is column E, username C, timestamp F, likes G, mentions K, hashtags L.

Comments mentioning a topic: =COUNTIF(E2:E5000, "*shipping*")
Topic share: divide that by =COUNTA(E2:E5000)
Questions: =COUNTIF(E2:E5000, "*~?*") — the tilde matters, because on its own ? is a wildcard in COUNTIF in both Excel and Google Sheets.
Comments per day: add a helper column =LEFT(F2, 10) and build a pivot table counting rows per date.
A topic over time: add a column flagging the topic (=ISNUMBER(SEARCH("shipping", E2))) and pivot it by date.
Your own categories: add a column and tag each comment by hand — praise, complaint, question, off-topic. For a few hundred comments this takes less time than people expect, and the result is more trustworthy than any automated label.

Step-by-step formula guides: Excel and Google Sheets.

When is AI sentiment analysis worth it?

To be fair to AI: language models can be genuinely useful for summarising thousands of free-text comments, grouping them into themes, or handling sarcasm better than a keyword ever will. If you are analysing tens of thousands of comments across many posts, or you need themes you didn’t think to search for, they can save real time.

But for a typical post, the trade-offs are worth knowing:

Short comments carry little signal. Many comments are an emoji, a tag or a row of heart-eyes. Scoring their sentiment adds a number without adding understanding.
Scores are hard to check. “68% positive” invites the question “according to what?”. Counts don’t have that problem.
Results can vary with the model, the prompt and the run, especially on slang, sarcasm and mixed languages.
Privacy. Sending comments to a third-party AI service is another place the data goes; make sure that fits your purpose and policies.

A sensible order is: count first, read the questions and most-liked comments, search the top words, and only then decide whether an AI pass would tell you something you don’t already know. If it would, the JSON or CSV export is ready to feed into whatever tool you choose.

Stats are free on every post. On posts over 200 comments the panel first shows the free preview, clearly labelled; a one-time Export Pass ($4.99 for up to 10,000 comments) gives stats on the whole post.

FAQ

Does IGComments use AI to analyze comments?

No. The stats panel is built from straightforward counts of what is in the comments — words, hashtags, mentions, question marks, emoji, likes and dates — so the same post always gives the same results.

Can I get sentiment analysis of Instagram comments?

IGComments doesn’t score sentiment. Top emoji and most-liked comments give a quick read of the mood, and you can export the comments as CSV or JSON to run them through any sentiment tool you choose.

Which languages does the word analysis support?

Words in any language are counted. Common filler words are removed in English, Spanish and Portuguese; in other languages some common words may appear in the list.

How are questions detected?

A comment counts as a question if it contains a question mark or an inverted question mark. It is simple and predictable, so it misses questions written without one.

Is the analysis free?

Yes, for every post. On posts with more than 200 comments, the stats first cover the free 200-comment preview until the full post is unlocked.